Can someone tell me which is the best turbulence (or even laminar) model for un incompressible flow (oil, density 850 kg/m3, Viscosity 30 cSt) and which the parametres for wall boundary (roughness model)?

Or if a best model doesn't exist, how can I treat the problem to get at least a good one?

If the flow is turbulent then personally I'd start with the simplist turbulence model and see what the results are like if it gives you what you expect then fine if not try the next more sophisticated one and so on.

The two-equation k-ep turbulence model is the most validated and its weaknesses/strengths are well documented.

allan